Hi,
On 6 Mar 2009, at 20:41, Vishwadeep Ahluwalia wrote:
> Hi,
> I want to find out the best Hrf model for orbitofrontal cortex to
> use for my group of
> subjects. I have 16s/32s on/off simple block design. I extracted the
> timecourse of a TICA
> melodic component that showed activation in OFC to use it in the GLM
> model. Should i
> have to change the PP height of the regressor derived from MELODIC
> to 1 (or same as the
> other models i'm comparing).
You may as well. Whilst rescaling the height to 1 won't change the
zstat on a contrast on this EV alone, it will affect contrasts between
this and other EVs, etc, so it's worth making it comparable.
> The second model is a 16s square wave convolved with
> gamma hrf. Looking at the model it seems that the time to peak is
> almost 16s. From the
> averaged Raw data BOLD plots , it seems that the BOLD peak is more
> like 6s. Because of
> these timing differences the model fit (and therefore PE's and z-
> scores) would be
> less..right??
Depends on the accuracy of the whole shape and the data - but if what
you're describing is the main differentiating feature then you may be
correct.
> What parameters in the model setup should i change to bring the peak
> forward in time? I tried changing the phase but that shifts the
> whole waveform forward.
> Should i change the square wave block length instead?
No - you should adjust the HRF parameters - for example in the Gamma
reduce the stddev and/or the mean lag.
> Another doubt i have is whether the GLM is more sensitive to height
> or width of the
> response or is it really the overall shape(area under the curve).
> i.e if i use a 10s square
> wave with gamma hrf to model one BOLD response with 1% signal change
> and duration
> of 12 s and a BOLD response with 0.5% signal change with duration of
> 16s , which will
> give me a higher PE value?
The stats on the GLM reflects the overall fit (i.e. sum of squared
differences between model and data at all timepoints) so any of these
factors could change that.
Cheers.
> Thanks for your patience and time
> -Vishwadeep Ahluwalia
>
---------------------------------------------------------------------------
Stephen M. Smith, Professor of Biomedical Engineering
Associate Director, Oxford University FMRIB Centre
FMRIB, JR Hospital, Headington, Oxford OX3 9DU, UK
+44 (0) 1865 222726 (fax 222717)
[log in to unmask] http://www.fmrib.ox.ac.uk/~steve
---------------------------------------------------------------------------
|